Publication number | US8024142 B1 |
Publication type | Grant |
Application number | US 11/961,543 |
Publication date | Sep 20, 2011 |
Filing date | Dec 20, 2007 |
Priority date | Dec 20, 2007 |
Fee status | Paid |
Publication number | 11961543, 961543, US 8024142 B1, US 8024142B1, US-B1-8024142, US8024142 B1, US8024142B1 |
Inventors | Mathieu Gagnon, Jurgen Hissen |
Original Assignee | Pmc-Sierra Us, Inc. |
Export Citation | BiBTeX, EndNote, RefMan |
Patent Citations (11), Classifications (6), Legal Events (6) | |
External Links: USPTO, USPTO Assignment, Espacenet | |
The present invention relates generally to characterizing digital communication circuits. More particularly, the present invention relates to analyzing received signal waveforms.
Current communication circuits have a high level of integration and achieve high performance by minimizing parasitic effects. During both product characterization and normal product usage, it is often desired to observe signal waveforms on critical high-speed internal nodes. However, current integrated circuits cannot be directly probed internally without significantly altering the signal to be observed.
Internal characterization is a very important, time-consuming and expensive part of the complete design cycle. At the current design speeds, sampling the signal internally is very difficult and requires dedicated output pads with linear buffers. These elements require significant area and their design is difficult. Even if turned off in normal operation, their presence can cause detrimental effects that limit the bandwidth and require improved performance from the circuits ahead. The cost of high-speed test equipment is also substantial.
Most existing methods for indirectly probing internal nodes use dedicated circuits to apply time and voltage offsets to probe different parts of the waveform.
A common way to represent a communication signal waveform is to plot its amplitude vs. the time elapsed from a fixed instant relative to the current bit period start. The graph resulting from showing many such waveforms collected over different bit periods is called an eye diagram. The bit error map described above has a shape similar to that of the eye diagram, as its probability distribution matches that of the signal traces around the “center opening” of the eye diagram.
Such previously known signal waveform analysis systems typically require a known pattern to be transmitted, in order to be able to compare the recovered signal with the expected value. This limits the application to systems where the transmitter is fully controllable, and can send one of the patterns that can be detected at the output. In a practical application, the data transmitted can be difficult to control and programming this configuration can require lengthy setup procedures. The analysis of critical internal node signal waveforms is also helpful to identify problems that arise when a specific pattern is transmitted. However, if the specific pattern is not one of the pre-defined test patterns, previously known analysis systems are not useful for this purpose.
In a first aspect, the present invention provides a method of analyzing a signal waveform. The method comprises digitally sampling a signal at a periodic sampling interval; accumulating a count of samples of the signal at a logic level relative to a threshold value over a given period; stepping the threshold value through a series of values and repeating the steps of digitally sampling and accumulating samples; and modifying the skew of the periodic sampling interval and repeating the steps of digitally sampling, accumulating the count of samples and stepping the threshold to determine a statistical distribution of the signal waveform. A signal density can also be calculated by determining the difference between the count of adjacent samples at successive threshold values for each sampling offset skew.
According to a further aspect, there is provided a system for analyzing a signal waveform. The system comprises a threshold control circuit to step a threshold value through a series of values; a sampler to sample a signal at a periodic sampling interval; a clock skew circuit to apply a clock offset to a clock signal to determine the periodic sampling interval and to step the periodic sampling interval through series of values; and an accumulator to accumulate, over a given period, a count of samples of the signal at a logic level relative to each of the series of threshold values, and at each of the series of periodic sampling intervals, to provide a statistical distribution of the signal waveform. A synchronization circuit controls the accumulator to accumulate a count of samples for the given period, and a clock recovery circuit recovers the clock signal.
In an embodiment of the system, a reference sampler provides reference samples sampled at a reference threshold and according to the clock signal, and a reference accumulator accumulates a count of the reference samples. The system can also include multiple sampler circuits operating sequentially, or multi-level sampler circuits simultaneously sampling at a plurality of threshold values.
Other aspects and features of the present invention will become apparent to those ordinarily skilled in the art upon review of the following description of specific embodiments of the invention in conjunction with the accompanying figures.
Embodiments of the present invention will now be described, by way of example only, with reference to the attached Figures, wherein:
Generally, the present invention provides a method and system for analyzing signal waveforms. The system can be integrated directly onto the device or chip, and can take advantage of existing on-chip components, such as samplers, comparators, and clock recovery circuits. The present system evaluates a signal waveform based upon its statistical distribution. It uses the pattern balance information contained in a digital signal, or the periodic nature of a signal to determine the cumulative statistical distribution of the signal, and it uses the difference between consecutive cumulative statistical distributions to evaluate the statistical density of the signal. A pre-defined bit sequence is not required.
The present invention provides both internal characterization and evaluation of the performance in high-speed communication systems. Existing circuitry is re-used and the extra functionality is pushed away from the sensitive high-speed parts of the communication system. For example, the extra circuits can be placed after the digital sample's frequency is reduced by digital means (e.g. after a serial-to-parallel interface). The overhead circuitry required to implement this functionality is thus small, easy to design, and does not interfere with the sensitive high-frequency nodes that are to be probed.
In most digital communication systems, the bits sent have an equal probability of being logic 1's or logic 0's over a sufficiently long period of time. This property ensures that an undistorted and noiseless signal sampled with a threshold selected to be between the logic 1 and logic 0 levels, at the appropriate time between the transitions, will return a number of 1's within a known range over a given period of time. Those skilled in the art will recognize that this range is determined by constraints imposed by the communication protocol used. Furthermore, by starting to count the number of bits based upon specific patterns meant for bit alignment, this range can sometimes be reduced.
For analyzing digital signals, the present system and method takes advantage of the limited range in the normal counts of logic 1's or 0's over a given period of time. If the signal is sampled at a threshold and instant within the bit period where the probability of getting a 0 or a 1 is equal, the final count will indicate a proportion of 1's close to 50%. When the threshold or sampling time is modified such that some bits that would normally sample as a logic 1 get sampled as a logic 0, the ratio of logic 1's will decrease below 50%. Thus, the expected distribution of 1's or 0's will vary as the sampling threshold and/or sampling instant gets closer to the location of the trace within the traces of the waveform eye diagram. The variation of the distribution of sampled 1's and 0's can then be used to deduce the signal waveform distribution over voltage and time skew relative to the bit period.
When sampling periodic waveforms repeatedly at a given instant in their period, the probability of getting a 1 will change as the sampler threshold is changed from well below the trace (where the probability of 1's will be close to 100%), to the average trace location (where the probability of 1's will be close to 50% due to noise), to well above the trace location (where the probability of 1's will be close to 0%). Thus, observed changes in the statistical distribution of a signal over the sampling voltage can be used to determine the distribution of the signal at different time instants within the period.
The output of the sampler 102 is a digital signal that represents whether the input signal was below or above the threshold at the beginning of the last sampling clock cycle. The value of this signal is digitally summed with previous values in an accumulator 116. In other words, the number of sampled 1's (or alternately 0's) output by the sampler 102 in a given duration of time at a given clock offset is counted by the accumulator 116. The time duration (i.e. starting instants and stopping instants) is controlled by a synchronization circuit 118. The synchronization circuit 118 can, for example, count a pre-determined number of bits or be synchronized with the incoming pattern to reduce the possible range of variation of the final count.
The final output 120 of the accumulator 116 is the count of the number of bits whose voltage was above (or below) the threshold value at the sampling instants, after a pre-defined number of bits is sampled. Those skilled in the art will recognize that there are many ways to have the accumulator 116 output this value. The synchronization circuit 118 can, for example, reset the accumulator 116 when the count is desired to start, and output the value of the accumulator when the desired number of bits has been received. It can also output the value of a continuously running accumulator both at the initial and final instants. In this case, the number of bits counted during the period is calculated from the difference between the final and initial accumulator values.
By recording the accumulation of 1's or 0's, as determined by the output 120 of accumulator 116, for equal periods of time over a range of thresholds 104 and clock skews 108, a 2-D map can be plotted. This map represents the probability that the incoming signal 56 is above the current threshold voltage at a given sampling instant within the bit period. An example of such a map is shown in
As can be seen, the contour lines 124 provide a good approximation of the eye diagram. The difference in shape between the lines 122 and the contour lines 124 can be explained by three factors: random noise is present in the true incoming signal 56; the variation of the threshold voltage and clock skew is performed in discrete steps (i.e. is discontinuous); and uncertainty introduced by variation in the number of logic 1's and 0's received during a given period of time, as determined by the specific communication protocol being used.
As described above, the maximal variation in the expected number of 1's or 0's over the period of time of the measurement can be bounded according to the protocol being used. The synchronization circuit 118 can also exploit specific properties of the incoming signal to synchronize the starting time of the accumulation, and its duration, so that this variation is minimized. The total number of bits accumulated, which corresponds to the time allowed for each accumulation, will usually determine the precision of the test. Since the variation of expected 1's or 0's is a fixed number, counting more bits will improve the precision.
In some protocols, if the output signal becomes too distorted for the synchronization circuit 118 to recognize it, a previously determined starting point can be used to adjust the start of the accumulation time. For example, in 8b10b protocols, the variation between the number of 1's and 0's is guaranteed to be either +1, 0 or −1, if the accumulation starts at the beginning of any 10-bit byte. When the location of the start of such a byte is determined from a previous run and the clock recovery is kept running, it is sufficient for the synchronization circuit 118 to start the next accumulation period any 10-bit multiple away from this known byte start to guarantee the same maximal mismatch of ±1 between the expected number of logic 1's or 0's received. Other characteristics of the signals can also be exploited to reduce this variation. Again, this is specific to the protocol being used.
As noted above, the output 120 of accumulator 116 for a given duration provides a number that is on average proportional to the probability that the recovered signal is above the threshold at a given sampling instant and sampling period. This can be explained by reference to the following equation:
X(threshold,skew)=P(signal(skew+n*period)>threshold)*N Eq.1
for n=0 . . . N−1; where:
The estimation x(threshold, skew) can then be used to determine the probability that the signal lies within two consecutive threshold voltages. That is:
Z(threshold,skew)=P(signal(skew+n*period)>threshold)AND signal(skew+n*period)≦threshold+dV)*N Eq. 2
for n=0 . . . N−1; where, in addition to the previously defined parameters:
To estimate the Z matrix, one can take the difference between the results at two adjacent voltage thresholds collected at the same clock skew. This is illustrated in
z(threshold,skew)=x(threshold,skew)−x(threshold+dV,skew) Eq.3
Thus, by calculating the difference between sampled bits accumulated at adjacent threshold settings, for a given clock skew, the probability that the signal is between a given threshold and (threshold+dV) at that instant can be derived. This provides a discrete approximation of the signal density over voltage and time.
A plot of the resulting matrix z(threshold,skew) can be seen in
z(ti,csi)=x(ti,csi)−x(ti+dV,csi)
by increasing ti by dV and csi by dt, as shown by the nested loops shown at steps 168, 170, 172, 174, 176 and 178. The resulting z matrix can then be post-processed and plotted (step 180), if desired, to arrive at a plot as shown in
Referring to Eq. 1 above,
X′(threshold,skew)=P(signal(skew+n*period)>threshold) Eq.4
for n=0 . . . N−1
An estimate of X′ (threshold, skew), termed x′ (threshold, skew) can then be estimated as:
x′(threshold,skew)=(accumulator count)/(reference accumulator count) Eq.5
x′ (threshold, skew) can then be used instead of x(threshold, skew) in Eq. 3 to derive the Z matrix.
It is also possible to evaluate the waveform of periodic sequences of bits (i.e. one that repeats over time) using the present system. Those skilled in the art will recognize that certain sequences, such as pulses or steps, can give important information about a system's characteristics. Periodic approximations of those sequences can be easily created. To analyze the signal waveform of such sequences, the system of
As will be apparent to those of skill in the art, this method can also be applied to non-periodic digital signals for which the reference clock frequency is faster than the incoming digital signal bit rate by an integer multiple. In this case, it can be seen as a way to extend the clock skew over more than one bit period.
This method can also be applied to a periodic, non-digital signal (i.e. an analog signal). In this case, the sampling clock will be generated from a fixed time base rather than from the input signal itself. The input signal period must be an integer multiple of the time base period.
As will be apparent to those of skill in the art, the present invention can be applied to half-rate or similar architectures. In this case, multiple samplers in parallel can each operate on different data samples, in sequence. The results of the sequential samplers can be combined, either in common accumulators, or through the summation of distinct accumulators. Similarly, multi-level samplers can be used, where each operates at the same time, but each with a different threshold value. Analog-to-digital converters are examples of such multi-level samplers.
As will be appreciated by those of skill in the art, the present system and method for analyzing a signal waveform provide many advantages over previously known systems and methods. The analysis circuitry can be included on the integrated chip, and can use many circuits and components normally required in communication systems, such as samplers, threshold control circuitry, and clock recovery circuitry.
The present invention can be applied to most high-speed communication systems. It can also be used to probe high-speed internal nodes, especially for microelectronic systems. The method is suitable for use in integrated circuit with minimal impact on the sensitive high-speed node to be probed. The system is also able to analyze signals of arbitrary information content—the receiver does not need to know in advance the content of the signal. As will be further appreciated, this permits assessment of the performance in final communication system in real-world situations. The present system and method can be used to identify problem causes in installed systems, as well as to predict quickly the noise immunity of such systems. The effect of various settings or configurations can be easily monitored and compared.
In the preceding description, for purposes of explanation, numerous details are set forth in order to provide a thorough understanding of the embodiments of the invention. However, it will be apparent to one skilled in the art that these specific details are not required in order to practice the invention. In other instances, well-known electrical structures and circuits are shown in block diagram form in order not to obscure the invention. For example, specific details are not provided as to whether the embodiments of the invention described herein are implemented as a software routine, hardware circuit, firmware, or a combination thereof.
The above-described embodiments of the invention are intended to be examples only. Alterations, modifications and variations can be effected to the particular embodiments by those of skill in the art without departing from the scope of the invention, which is defined solely by the claims appended hereto.
Cited Patent | Filing date | Publication date | Applicant | Title |
---|---|---|---|---|
US4386526 * | Mar 26, 1981 | Jun 7, 1983 | Eckhard Roeder | Method for quality control of processes and construction components |
US6463109 | Mar 20, 2000 | Oct 8, 2002 | Vitesse Semiconductor Corporation | Multiple channel adaptive data recovery system |
US6694462 | Aug 9, 2000 | Feb 17, 2004 | Teradyne, Inc. | Capturing and evaluating high speed data streams |
US6920402 | Mar 7, 2001 | Jul 19, 2005 | Rambus Inc. | Technique for determining performance characteristics of electronic devices and systems |
US7137048 | Oct 12, 2001 | Nov 14, 2006 | Rambus Inc. | Method and apparatus for evaluating and optimizing a signaling system |
US7143323 | Dec 13, 2002 | Nov 28, 2006 | Teradyne, Inc. | High speed capture and averaging of serial data by asynchronous periodic sampling |
US7185252 | Mar 8, 2005 | Feb 27, 2007 | Samsung Electronics Co., Ltd. | Measurement circuit and method for serially merging single-ended signals to analyze them |
US7194365 | Aug 31, 2004 | Mar 20, 2007 | Synopsys, Inc. | Method and apparatus for integrated undersampling |
US7262621 | Mar 21, 2005 | Aug 28, 2007 | Synopsys, Inc. | Method and apparatus for integrated mixed-signal or analog testing |
US20030169755 * | Mar 11, 2002 | Sep 11, 2003 | Globespanvirata Incorporated | Clock skew compensation for a jitter buffer |
US20040019458 | Apr 30, 2003 | Jan 29, 2004 | Samsung Electronics Co., Ltd. | Jitter measuring system in high speed data output device and total jitter measuring method |
U.S. Classification | 702/69 |
International Classification | G06F17/18 |
Cooperative Classification | G01R13/0272, G01R19/2509 |
European Classification | G01R19/25C2, G01R13/02C6 |
Date | Code | Event | Description |
---|---|---|---|
Dec 21, 2007 | AS | Assignment | Owner name: PMC-SIERRA, INC., CALIFORNIA Free format text: ASSIGNMENT OF ASSIGNORS INTEREST;ASSIGNORS:GAGNON, MATHIEU;HISSEN, JURGEN;REEL/FRAME:020282/0252 Effective date: 20071219 |
Aug 6, 2013 | AS | Assignment | Owner name: BANK OF AMERICA, N.A., NORTH CAROLINA Free format text: SECURITY INTEREST IN PATENTS;ASSIGNORS:PMC-SIERRA, INC.;PMC-SIERRA US, INC.;WINTEGRA, INC.;REEL/FRAME:030947/0710 Effective date: 20130802 |
Mar 4, 2015 | FPAY | Fee payment | Year of fee payment: 4 |
Feb 1, 2016 | AS | Assignment | Owner name: PMC-SIERRA US, INC., CALIFORNIA Free format text: RELEASE BY SECURED PARTY;ASSIGNOR:BANK OF AMERICA, N.A.;REEL/FRAME:037675/0129 Effective date: 20160115 Owner name: PMC-SIERRA, INC., CALIFORNIA Free format text: RELEASE BY SECURED PARTY;ASSIGNOR:BANK OF AMERICA, N.A.;REEL/FRAME:037675/0129 Effective date: 20160115 Owner name: WINTEGRA, INC., CALIFORNIA Free format text: RELEASE BY SECURED PARTY;ASSIGNOR:BANK OF AMERICA, N.A.;REEL/FRAME:037675/0129 Effective date: 20160115 |
Feb 3, 2016 | AS | Assignment | Owner name: MORGAN STANLEY SENIOR FUNDING, INC., NEW YORK Free format text: PATENT SECURITY AGREEMENT;ASSIGNORS:MICROSEMI STORAGE SOLUTIONS, INC. (F/K/A PMC-SIERRA, INC.);MICROSEMI STORAGE SOLUTIONS (U.S.), INC. (F/K/A PMC-SIERRA US, INC.);REEL/FRAME:037689/0719 Effective date: 20160115 |
Mar 22, 2016 | AS | Assignment | Owner name: MICROSEMI STORAGE SOLUTIONS, INC., CALIFORNIA Free format text: CHANGE OF NAME;ASSIGNOR:PMC-SIERRA, INC.;REEL/FRAME:038067/0428 Effective date: 20160115 |